Autonomous sensor networks

The capacity of a radio channel as studied by Shannon in 1948 is known, and solutions exist to achieve it. The question then arises as to how much attention should be paid to the physical layer and hardware architecture of communication systems. However, the evolution of sensor networks towards the Internet of Things is creating the need for a new understanding of these systems, and bringing with it a very strong constraint on energy consumption.

In recent years, the first large-scale sensor networks have been deployed. While the first sensor network dates back to 1969 (seismic sensors used by the US army in Vietnam), there are now almost 10 billion communicating objects (not just sensors), and the scope of applications suggests that this number is set to grow rapidly: intelligent cities and buildings, vehicles, natural disasters and rescue, autonomy for the elderly, inventories, etc. Figure 1 illustrates an application of sensor networks dedicated to monitoring the ageing of infrastructures.

Wireless sensor networks enable data to be collected, processed locally or transmitted to management centers. But communications constraints differ from what we've been doing until now:

• the lifetime of the nodes must be long enough to ensure a sufficiently long network life (years, decades) without having to change the batteries;

• network densification requires a high degree of robustness, particularly to interference whose probability density does not necessarily have the classic Gaussian appearance.